Measuring High-Energy -Ray Spectra with HAWC Sam Marinelli for the - PowerPoint PPT Presentation

Measuring High-Energy γ -Ray Spectra with HAWC Sam Marinelli for the HAWC Collaboration Michigan State University August 9, 2017 S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 1 / 13

The High-Altitude Water-Cherenkov observatory Detects TeV γ rays at 4100 m on the Sierra Negra mountain in Puebla, Mexico. 1200 PMTs in 300 water-filled tanks detect Cherenkov light from air showers. Timing used to determine shower direction. S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 2 / 13

HAWC tanks S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 3 / 13

HAWC energy estimation via artificial neural network (NN) Input Layer Output Layer Hidden Layer Using Toolkit for w 1 y 2 Multivariate Analysis 1 . 11 1 x y 1 w 1 1 w 2 1 12 11 NN maps several y 2 2 event-wise variables to y 1 x 2 2 y 2 y 3 y estimate of primary 3 1 ANN x y 1 3 energy. 3 y 2 4 479 free parameters x y 1 w 2 4 4 w 1 51 45 chosen by training on y 2 5 w 2 w 1 1 01 Bias 05 Monte Carlo (MC). Bias 1 1 http://tmva.sourceforge.net/ . S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 4 / 13

NN input variables Input variables chosen to characterize shower size and geometry. Shower characteristic Input variables Energy deposited in the detec- Fraction of PMTs hit tor Fraction of tanks hit Normalization of lateral-distribution fit Fraction of ground energy Distance of shower core landing on the detector from detector center Fraction of primary energy Zenith angle reaching the ground Lateral energy distribution S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 5 / 13

Performance on simulation NN energy better correlated with MC truth than currently used variable (fraction of PMTs hit). Ability to determine energies beyond 100 TeV. NN energy Fraction of PMTs hit 15 0 − 2 − 10 2 10 14.5 − 0.2 [Reconstructed energy (eV)] − 3 − 14 10 3 10 13.5 − 10 4 Fraction of Events − − 0.4 10 4 Fraction of Events 13 − hit 5 − 10 5 10 f 12.5 − 0.6 10 log − 6 − 12 10 6 10 − 0.8 11.5 − 10 7 − 10 7 10 log 11 − 1 − 8 − 10 8 10 10.5 − 9 − − 10 10 9 1.2 10 10 10.5 11 11.5 12 12.5 13 13.5 14 14.5 15 10 10.5 11 11.5 12 12.5 13 13.5 14 14.5 15 log [True energy (eV)] log [True energy (eV)] 10 10 S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 6 / 13

RMS error RMS error of ∼ 32% at highest energies. Use of lateral distribution compensates for fluctuations in height of first interaction. 0.8 Other techniques Neural Net Likelihood Ground Parameter 0.7 Ground 0.6 Parameter Log-energy RMS error 0.5 energy- 0.4 reconstruction ~100% method described 0.3 ~80% ~60% by Kelly Malone 0.2 ~40% on August 8 at 0.1 ~20% 15:00. 0 11 11.5 12 12.5 13 13.5 14 14.5 log [True energy (eV)] 10 S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 7 / 13

HAWC Crab spectrum using NN Events binned two-dimensionally in fraction of PMTs hit and NN energy. Poisson-likelihood forward-folded fit is applied to these bin contents. Crab modeled as point source with log-parabola γ -ray spectrum: dN dE = k ( E / E 0 ) − α − β ln( E / E 0 ) . (1) Fit serves as proof of principle for energy reconstruction but may also constrain high-energy Crab physics. S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 8 / 13

Crab fit result Statistical errors using 10 -10 new energy variables are Preliminary smaller than in published HAWC result 2 . 10 -11 E 2 dN/dE (TeV/cm 2 /s) Systematics analysis in progress. Assuming 50% 10 -12 flux systematic from published analysis, fits Neural Net Ground Parameter with new energy H.E.S.S. ICRC 2017 10 -13 10 0 10 1 10 2 variables are compatible E (TeV) with H.E.S.S. Dark band – statistical error measurement. Light band – systematic error 2 https://arxiv.org/abs/1701.01778 . S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 9 / 13

HEGRA Crab Nebula spectrum (Aharonian et al. 2014) Stat. errors at highest energies comparable to HEGRA’s. Might be improved with tuned cuts. S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 10 / 13

Implications of measurement for PWN modeling Interpretation of HAWC result requires understanding at what energies spectrum is being measured. High-energy γ spectrum sensitive to highest-energy electron acceleration. Models De Jager et al. model PWN high-energy inverse-Compton emission. Atoyan and Aharonian (1995) also suggest bremsstrahlung could play a role if PWN inhomogeneous. De Jager et al. (1996). S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 11 / 13

Sensitivity to Lorentz-invariance violation Lorentz- invariance- violating models predict γ decay to e + e − above some energy. Detection of high-energy γ rays constrains this energy scale. HAWC Crab spectrum will Mart´ ınez-Huerta and P´ erez-Lorenzana (2017). imply some limit. S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 12 / 13

Future work Crab analysis not yet optimized. Must tune cuts etc. to new spectral-fitting technique. Galactic plane in 56–100 TeV map, made with 1 ◦ extended-source model and assuming 2.63 spectral index, shows several known sources. With new energy variables, HAWC can attempt measurements of these sources’ spectra at unprecedented energies. PRELIMINARY 2 1 b [ ◦ ] 0 1 2 76 75 74 73 72 71 70 69 68 67 66 65 64 63 62 61 60 59 58 57 56 55 54 53 52 51 50 49 48 47 46 45 44 43 42 41 40 39 38 37 36 35 34 33 32 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 l [ ◦ ] 0 1 2 3 4 5 6 Significance S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 13 / 13

Bonus round Backup slides. S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 14 / 13

Sensitivity to time variability Mart´ ın et al. numerically models time dependence of spinning down pulsar/PWN. Cooling time for PeV electrons is ∼ 1 month. HAWC could look for spectral variations on this Mart´ ın et al. (2012). time scale. S. S. Marinelli (MSU) High-Energy Spectra with HAWC August 9, 2017 15 / 13